University of Wollongong Research Online

Faculty of Engineering and Information Faculty of Engineering - Papers (Archive) Sciences

1-1-2009

Distribution of impurities in via silicothermic reduction

Winny Wulandari Swinburne University of Technology

Akbar Rhamdhani Swinburne University of Technology

Geoff Brooks Swinburne University of Technology

Brian J. Monaghan University of Wollongong, [email protected]

Follow this and additional works at: https://ro.uow.edu.au/engpapers

Part of the Engineering Commons https://ro.uow.edu.au/engpapers/1242

Recommended Citation Wulandari, Winny; Rhamdhani, Akbar; Brooks, Geoff; and Monaghan, Brian J.: Distribution of impurities in magnesium via silicothermic reduction 2009, 1401-1415. https://ro.uow.edu.au/engpapers/1242

Research Online is the open access institutional repository for the University of Wollongong. For further information contact the UOW Library: [email protected] Distribution of impurities in magnesium production via silicothermic reduction

Distribution of impurities in magnesium via silicothermic reduction

Winny Wulandari1, M. Akbar Rhamdhani1, Geoffrey Brooks1, and Brian J. Monaghan2 1Faculty of Engineering and Industrial Sciences, Swinburne University of Technology John St, Hawthorn, Melbourne VIC, 3122, Australia 2Faculty of Engineering, University of Wollongong Northfield Ave, Wollongong NSW, 2522, Australia

Abstract

The desire for light weight materials, particularly in the automotive field, is fuelling greater magne- sium production. The Pidgeon process is currently the most widely used process for the production of magnesium. This batch process involves reduction of dolomite by ferro-, carried out at temperature between 1100-1200oC under vacuum in a retort, producing magnesium vapour which is then cooled and collected as a condensate. The major attractions of the process are its simplicity and low capital cost; however, the process is also labour and energy intensive. The Magnetherm process is also a vacuum batch process based on similar chemistry but operates at higher tempera- tures with magnesia dissolved in a slag. Various operational issues relating to the batch and vacuum nature of the process have limited its development. Higher temperatures will improve the kinetics of the silicothermic reduction but will also promote greater impurities into the vapour. Thermody- namic analysis of the distribution of impurities during condensation of magnesium vapour has been carried using FactSage thermodynamic package. The results from the multi-stage equilibrium model show that at equilibrium it should be possible to condense solid impurities before the major- ity of the magnesium vapour is condensed below 750oC, depending on the operating pressure of the system and the particular chemistry of the process. The model predicts that purities above 99.8wt% can be achieved for the Pidgeon process and 99.4wt% for the Mintek process. Without multi-stage condensation, the impurities in the metal are predicted to be significantly higher for the Mintek process (2.14wt% compared to 0.6 wt%). These results are broadly consistent with existing indus- trial data.

1 Introduction

Magnesium has a density of 1700 kg/m3, which is two-thirds and one-sixth of the density of alumin- ium and steel, respectively. Its wide application in automotive and construction purposes has pro- moted an increase in magnesium consumption in recent years [1]. The dominant process to extract magnesium from its ores is the Pidgeon process where the magnesium is reduced from dolomite

Proceedings of EMC 2009 1401 Wulandari, Rhamdhani, Brooks, and Monaghan through silicothermic reaction at moderate temperature. The process has dominated the production of magnesium in the world the last ten years with the proportion of almost 70% of the total magne- sium produced [2]. The main advantage of this process is the simplicity of the process and the high purity of the magnesium produced. Other silicothermic processes include the Magnetherm and the Mintek process. Table 1 shows the differences between the three different silicothermic processes, in terms of operating conditions and the level of impurities produced. The higher temperature and atmospheric pressure as in the Mintek process operating condition lead to higher productivity of magnesium, but also leads to lower purity magnesium metal. Table 1: Silicothermic process operating condition and impurities

Pidgeon [3, 4] Magnetherm [5-7] Mintek [8, 9] Operating Conditions Pressure 10 – 67 Pa 5 kPa 0.1 MPa Temperature (oC) 1100 – 1200 1550 – 1600 1700 – 1750 Productivity per 50 kg 20 ton 100 ton furnace per day Indicated impurities in crude magnesium, (in wt%) Al 0.004 0.01 0.066 Si 0.010 0.05 0.281 Ca 0.005 0.005 0.385 Fe 0.007 0.005 0.250 In this paper, the process production of magnesium metal through silicothermic processes and the associated impurities in magnesium metal will be reviewed. There is only limited work published on the thermodynamics of magnesium condensation [10-12]. The work of Cameron et al. [11] and Brooks et al. [12] focused on the carbothermic route and not the silicothermic route. In this study, we calculated the composition of condensates using Gibbs energy minimisation to predict the spe- cies and compounds in the system at equilibrium, as a means of analysing the potential for produc- ing high purity magnesium from a silicothermic process.

2 Overview of silicothermic processes

Silicothermic process involves a reduction reaction of magnesia or dolomite by silicon in the form of ferrosilicon to produce magnesium vapour which is then condensed in a condenser as magnesium solid metal. The detail of the processes, focusing on the thermochemistry of each process is ex- plained in the following sub-sections.

2.1 The Pidgeon process

The raw materials for the Pidgeon process are calcined dolomite (CaO.MgO) and ferrosilicon (Fe- 75wt% Si). The calcined dolomite and ferrosilicon are mixed and briquetted to improve the rates of heat transfer and the solid-state reaction. The briquettes are located inside a horizontal Ni-Cr

1402 Proceedings of EMC 2009 Distribution of impurities in magnesium production via silicothermic reduction stainless-steel retort as shown in Figure 1 and the reduction is carried out at the operating condi- tions. The reaction of the Pidgeon process can be written as:

. 2 CaO MgO(s) + (x Fe) Si(s) = 2 Mg(s) + 2 CaO.SiO2(s) (1) The magnesium vapour evolved from Reaction (1) is condensed on a removable sleeve in the throat of the retort [4]. Magnesium deposit is collected as a dense “crown” [3].

Figure 1: Schematic of the Pidgeon process retort [13] Table 2: Reported impurities of the Magnesium produced from the Pidgeon process (in wt%)

Impurities Laboratory scale [14] Pilot Plant Scale [3] Plant Scale [4]

Fe 0.002 0.003 Ni 0.002 < 0.001 < 0.005 Cu 0.0002 0.002 0.005 Si 0.07 0.003 0.008 Al 0.02 < 0.007 Mn < 0.001 0.001 Na < 0.0005 < 0.005 Pb 0.06 Ca 0.003 Sn 0.002 Zn 0.03 Pidgeon and King [15] has measured the equilibrium vapour pressure of magnesium in Reaction (1) and reported a value of 1346 Pa at 1100oC. The magnesium vapour produced in the reaction zone is removed by applying low pressures of the order of 6 to 67 Pa. At these operating conditions, the major species in gas phase is magnesium; while the other species such as calcium oxide, magne- sium oxide, silicon, and dicalcium silicates, have very low concentrations. Table 2 provides data on the impurities of magnesium metal produced via the Pidgeon process. Some metal impurities such as Ni, Cu, Pb, Sn, Mn, and Zn come from volatilization of the oxide impurities in dolomite [3]. The amount and the composition of these impurities depend on the com- position of starting materials. It can be seen that the Pidgeon process can produce magnesium with a very low level of impurity (>99.9 wt%). Post-oxidation of magnesium is often the major concern with impurities and not the intermetallic elements [16]. However, this process is labour intensive

Proceedings of EMC 2009 1403 Wulandari, Rhamdhani, Brooks, and Monaghan and has low productivity [17] and there has been significant recent research on developing new process routes to magnesium [9, 12, 18].

2.2 The Magnetherm process

The development of thermal reduction of magnesium (which leads to the development of the Mag- netherm process) began in in 1948 [5] and continued production into the 1990s [19]. How- ever, at present no industrial plant is operational. The process differs from the Pidgeon process in a number of ways [20]. The reduction is carried out in a liquid state, i.e. from a molten slag bath. Alumina/bauxite is used as a flux to form silico-aluminate slag, and the magnesium vapour pro- duced is condensed to liquid state, before it is sent for refining. The main starting materials for pro- duction of magnesium at the Beaudeant plant [5] using this process were calcined dolomite, ferro- silicon, and bauxite. The Magnetherm process involves a silicothermic reduction of MgO from the calcined dolomite (CaO.MgO) from a molten slag bath according the following overall reaction:

. . 2 CaO MgO + (x Fe) Si + n Al2O3 = 2 CaO SiO2.nAl2O3 + 2 Mg + x Fe (2)

The process is operated at temperature of 1550oC and pressure of 5 kPa inside an electric arc fur- nace, as shown in Figure 2. The slag composition is maintained at 55 wt% of CaO, 25 wt% of SiO2,

14 wt% of Al2O3, and 6 wt% of MgO. At the operating temperature, the slag system is not fully molten, with 40% in the form of solid of dicalcium silicate (2CaO.SiO2) [6]. The simplified silicothermic reduction can be written as:

(MgO) + Si = (SiO2) + 2 Mg(g) (3) where ( ) refers to the component being dissolved in the molten slag. The key to this process is to promote a low silica activity in the liquid slag and maintain the composition of the slag on the di- calcium silicate and MgO periclase phase boundary in the quaternary CaO-Al2O3-SiO2-MgO sys- tem [6]. The liquid is kept at fully saturated, hence the activity of MgO is nearly unity. The vacuum condition in the reactor results in a number of operational problems. The ingress of air during slag tapping was reported to cause the loss of 20% of the production due to the formation of

MgO and Mg3N2 in the condenser [7]. Another problem also comes from the oxidation of carbon from the reactor lining and electrode to form carbon monoxide, according to following reaction:

(MgO) + C(s) = Mg(g) + CO(g) (4)

The direction of the Reaction (3) is reversed at lower temperature. Hence, at low temperature, the carbon monoxide vapour produced and taken along the magnesium vapour stream will re-oxidise the magnesium vapour forming magnesium oxide. The magnesium vapour produced from the reaction is then condensed to liquid form in a condenser when it is cooled to a solid and sent for refining. Table 3 shows the impurities in the magnesium

1404 Proceedings of EMC 2009 Distribution of impurities in magnesium production via silicothermic reduction after refining process as reported by Bowman [21]. It is suggested that the impurities in the magne- sium originated from: (1) particulate matters that are carried over by magnesium vapour stream and condense in magnesium crucible as impurities, such as MgO and CaO; (2) physical and chemical interactions between species inside reactor; and (3) vaporization of volatilized materials, such as Mn, Zn, Pb, and Ni. Table 3: Impurities in the magnesium produced from the Magnetherm process before and after refining (using MgCl2 and KCl fluxes) process, in wt% [21] Impurities Initial After refining Calcium 0.77 – 1.05 0.005 Silicon 0.11 – 0.16 0.063 – 0.11 Aluminium 0.037 – 0.088 < 0.05 The details of the refining practice has been described by Bowman [21]. In this process the crude magnesium is remelted and flux containing 45% of MgCl2 and 55% of KCl is added and stirred into the melt. At the end of the refining process, the fluxes are allowed to settle before molten magne- sium is ladled and taken into casting. The chemical reaction in this refining stage can be written as:

MgCl2 + Ca = Mg + CaCl2 (5)

Calcium present in a metallic form is consumed by the flux. The flux also cap- tures oxide inclusion and form a sludge of heavy oxy-chloride [22]. The refining practice can produce a magnesium ingot with a standard composition as required by ASTM B92. However, the refining stage means a higher production cost and less magnesium metal yield. There is about 5-8% magnesium loss in the refining stage [21]. Another silicothermic process that has been developed in Italy and Brazil is called the Bolzano process[23]. It employs an internally heated and brick-lined reactor. The briquette charge contain- ing calcined dolomite and silicon is loaded inside the furnace. The process operates at 1200oC and a vacuum pressure at less than 400 Pa. Magnesium with a purity of 99.98-99.99% is obtained.

2.3 The Mintek process

There has been previous work by Cameron and co-workers and Christini on developing an atmos- pheric silicothermic route to magnesium [7, 19, 24], as a means of avoiding the difficulties associ- ated with Magnetherm and productivity limitations of the Pidgeon process. Currently, Mintek are developing an atmospheric process based on Schoukens’s patent [25]. The schematic of the Mintek process furnace is also shown in Figure 2. The key to this process relies on the utilization of alu- minium in addition to ferrosilicon for the reduction of magnesium oxide at higher temperatures, 1700-1750oC. The process also utilizes slag that remains liquid in the temperature range. Thus, the magnesium vapour produced is much higher, at about 0.85 atm. Argon gas is used to maintain the atmospheric pressure inside the furnace. Aluminium is more expensive compared to ferrosilicon.

Proceedings of EMC 2009 1405 Wulandari, Rhamdhani, Brooks, and Monaghan

The work is still underway for finding the proper feed composition for the optimum production of magnesium [22].

Figure 2: Schematic of the Magnetherm (left) [23] and Mintek (right) [26] processes The main advantage of the Mintek process is its high productivity. It also operates at atmospheric pressure which gives a possibility of operating the process in a continuous way. The magnesium produced, however, has very high impurities compared to the other two previous processes, as shown in Table 1. For example, the calcium concentration (main impurity) in the magnesium is 0.385 wt%, which is much higher compared to calcium concentration in the magnesium produced by the Pidgeon process, i.e. 0.005%. The magnesium metal produced need to be remelted and fur- ther refined. Table 4 shows the impurities in the magnesium before and after refining stage. The total impurities concentration in the magnesium before refining may exceed 2 wt%. The refining practice in the Mintek process consists of two stages [22]. In the first stage the magne- sium is refined by adding MgCl2-KCl flux to remove the calcium impurity, as per Equation (5). This stage is similar to the refining practice in the Magnetherm process. In the second stage, the molten magnesium is treated by adding FeCl3 to remove silicon impurity based on the following reaction:

Si + 1.5 Mg + FeCl3 = FeSi + 1.5 MgCl2 (6) Table 4: Impurities in the magnesium produced from the Mintek process before and after refin- ing process, in (wt%) [22] ASTM B92 Crude Mg Clean Mg Element Grade 9980A Avg. Max Min Avg. Max Min Aluminium 0.05 max 0.066 0.210 0.010 0.040 0.235 0.003 Silicon 0.05 max 0.281 1.010 0.100 0.095 0.500 0.014 Calcium 0.05 max 0.385 1.090 0.030 0.023 0.102 0.005 Iron 0.05 max 0.250 0.350 0.002 0.047 0.100 0.002

1406 Proceedings of EMC 2009 Distribution of impurities in magnesium production via silicothermic reduction

Abdellatif reported that the loss of magnesium in the refining stage in the Mintek process is about 9 % higher than that of the Magnetherm refining process, which ranges from 5 to 8% [22].

3 Thermodynamic analysis of silicothermic processes

There have been a number of studies on the thermodynamic analysis of the Pidgeon, the Magneth- erm, and the Magnetherm atmospheric processes [6, 19, 24, 27]. However, there is limited informa- tion on the literature that focuses on the behaviour of the impurities present in magnesium. Knowl- edge on the distribution of the minor phases and impurities and their behaviour is important for the improvement of these processes. In the present study, a thermodynamic analyses and calculations of the silicothermic processes, with an emphasis on the impurities distribution, were carried out using the FactSage thermodynamic package (version 5.0) [28]. FactSage is an integrated database computing system for chemical thermodynamic. This package has optimized database for solutions, such as metals, liquid and solid oxides, and slags. For pure components, the data are taken mostly from JANAF Thermochemical Tables [29], thermodynamic properties data from Barin [30], and from Berman et.al. [31]. The solu- tion for liquid slag phase is taken from Erikkson et.al [32] using a modified quasi-chemical model for the oxide. In Equilib module, the Gibbs energy minimization technique is used to calculate the concentration of chemical species when specified elements or compounds are react to reach the state of equilibrium. Phases from the compound and solution databases are retrieved and listed as possible product in the equilibrium result. The appropriate optimised thermodynamic solution data- bases were used in the modelling, which include the solution databases for monoxide, liquid slag, a and a’-dicalcium silicate. In the current modelling, magnesium solid is considered as a single phase, i.e. no solid solution is considered inside the magnesium. In future work, a condensed metal- lic solution phase will be included in the calculation. For the Pidgeon process equilibrium calculation, the input data including the calcined dolomite composition and the ratio of calcined dolomite to ferrosilicon were taken from Toguri’s data [14] and is listed in Figure 4. The addition of CaF2 as catalyst as reported in the literatures [3, 14] was neglected as the catalyst is only affecting the kinetics but do not change the equilibrium. The equi- librium was evaluated at temperature of 1160oC and pressure of 67 Pa, replicating the practical op- eration conditions [3, 14]. The input data for the Magnetherm equilibrium calculation were taken from Faure and Marchal [5] and Christini, and is listed in Figure 5 [6]. The liquid slag solution da- tabase and alpha dicalcium silicate (α-Ca2SiO4) phase were included in the equilibrium calculation. The equilibrium was calculated at the operating conditions of the Magnetherm process, i.e. at tem- perature of 1550oC and pressure of 5 kPa. For equilibrium calculation of the Mintek process, the input data for the calcined dolomite, ferrosilicon, and aluminium were taken from one of the com- positions from small-plant scale trials as reported by Abdellatif and listed in Figure 6 [26]. The cal- culation equilibrium was carried out at temperature and pressure of 1750oC and 85 kPa, respec- tively.

Proceedings of EMC 2009 1407 Wulandari, Rhamdhani, Brooks, and Monaghan

Further thermodynamic modelling was also undertaken to study the precipitation of solid impurities from the vapour phases produced upon cooling by carrying out a multi-stage equilibrium modelling using the same thermodynamic package. Figure 3 illustrates the multi-stage equilibrium modelling carried out in the present study. In Stage 1, the equilibrium at 1100oC was evaluated. The solid phases/impurities formed were then removed from the system and the vapour phases produced were cooled down to 1050oC; and the Stage 2 equilibrium calculation was then carried out. The proce- dure was repeated for the equilibrium calculations in the subsequent stages (where at each stage the temperature is decreased by 50oC) until the magnesium metal was condensed.

Stage 1 Stage 2 Stage .. Vapour vapour vapour vapour 1100oC 1050oC 550oC o o o … o

solid impurities solid impurities magnesium metal

Figure 3: Schematic of a multi-stage equilibrium modelling carried out in the present study

4 Results and discussion

The one-stage equilibrium model was used to determine the possible phases and its composition in the equilibrium. Figure 4 shows the results of the equilibrium calculation for the Pidgeon process at 1160oC and 67 Pa. For this process, the solid phases at equilibrium were alpha prime dicalcium silicate (α’-Ca2SiO4); monoxides which include single phase periclase (MgO) and calcined dolo- . mite (CaO MgO) containing Al2O3; ferrosilicon; and calcium aluminate. The vapour phase con- tained 99.68% Mg, and impurities which were mainly Ca and Fe. The presence of SiO vapour at relatively moderate temperature was very low, as suggested by Toguri and Pidgeon [33]. The va- pour phase was then condensed to 550oC, a temperature that is lower than its melting point, i.e. 650oC. The pressure of the system was below the vapour pressure of Mg, thus it condensed to solid, as stated by Betcherman and Pidgeon [34]. At this temperature, the model predicted that the solid contain 99.32 wt% Mg. It also predicted that other elements end up as impurities, with the majority in the form of CaMg2 intermetallic. In reality, the calcium is likely to be dissolved in the Mg, an aspect of the process not considered in this current model. The equilibrium temperature of condensation of Mg under vacuum predicted in this model was in a good agreement with a previous study by Misra et al. [35], on the condensation characteristic of magnesium vapour from the Pidgeon process. According to their experimental study, the Mg metal condensed at temperatures below 590oC. At 470oC, the recovered deposit was coarse, while below 200oC it was very fine and pyrophoric.

The results of the equilibrium calculation for the Magnetherm process at 1550oC and 5 kPa are pre- sented in Figure 5. In the Magnetherm process, the input charge includes dolomite, ferrosilicon and

1408 Proceedings of EMC 2009 Distribution of impurities in magnesium production via silicothermic reduction bauxite. Slag in the reactor also involved in the reaction thus it is included in the input for the equi- librium calculation. At equilibrium, the predicted non-vapour phases present include liquid slag, alpha dicalcium silicate, and ferrosilicon. The predicted vapour phase contains 98.94% of Mg and other impurities which include Ca, Fe, and Al. The presence of SiO vapour predicted is significant compared to the case of Pidgeon process, i.e. 0.18% of the total vapour phase. The vapour phase was brought down to 650oC, i.e. temperature at the condenser. At this condition, the Mg was in a liquid form, 99.55% of the total material. The major impurity is Ca in the form of inter-metallic

CaMg2, which is similar to the case of the result from the calculation for the Pidgeon process.

Dolime 1000 kg Condensation of Vapour 197 kg SiO2 0.42% o FeO 1.34% Mg 99.68% magnesium at 550 C Silicothermic reduction (g) MgO 36.13% Ca (g) 0.31% CaO 61.74% o at 1160 C and 67 Pa Fe (g) 0.01% Al2O3 0.37% SiO(g) 18.5 ppm α’ - Ca2SiO4 1447 kg Solid 197 kg Al (g) 0.5 ppm FeSi 194 kg Ca2SiO4 99.58% Mg 99.32% Mg2(g) 0.4 ppm Fe 25% Mg2SiO4 4.42% CaMg2 0.67% Si 75% Fe_bcc 0.01% Oxide 129 kg MgO 36.13% Ca2Si 46.4 ppm CaO 61.74% CaO 25.8 ppm

Al2O3 0.37% CaAl2 0.9 ppm MgO periclase 19 kg FeSi 89 kg

3CaO.Al2O3 9 kg

Figure 4: Equilibrium calculation for the Pidgeon process at 1160oC and 67 Pa.

Dolime 1000 kg Condensation of SiO 0.23% Vapour 201 kg 2 magnesium at 650oC FeO 0.42% Mg(g) 98.94% MgO 38.88% Silicothermic reduction Ca (g) 0.82% CaO 60.40% at 1550oC and 5 kPa SiO(g) 0.18% Bauxite 158 kg Fe (g) 0.05% Al2O3 100% Metal 201 kg Al (g) 23.9 ppm α - Ca2SiO4 807 kg Mg (l) 98.55% FeSi 189 Mg2(g) 16.2 ppm Ca2SiO4 94.65% CaMg 0.72% kg Al O 2.34 ppm 2 Fe 25% Mg2SiO4 5.35% 2 (g) Fe_bcc 0.05% Si 0.51 ppm (g) Ca Si 0.44% Slag 338 kg Slag 1190 kg 2 CaO 0.23% SiO2 25% MgO 6.02% Al2O3 14% CaO 43.11% CaAl2 44.7 ppm MgO 6% Al2O3 35.87% CaO 55% SiO2 15.00% FeSi 76 kg

Figure 5: Equilibrium calculation for the Magnetherm process at 1550oC and 5 kPa.

Proceedings of EMC 2009 1409 Wulandari, Rhamdhani, Brooks, and Monaghan

The results of the equilibrium calculation for the Mintek process are shown in Figure 6. The pre- dicted vapour phase contains 2.14 wt% of impurities, consisting of mainly Ca, followed by Mn and SiO vapour. For the Mintek process, the vapour is predicted to condense at 750oC and the Mg con- sists of 96.16% of the condensed phases. Similar to the other processes the major impurity pre- dicted in this model is in the form of CaMg2.

Dolime 1000 kg Condensation of SiO 0.98% Vapour 227 kg 2 o Fe O 0.31% magnesium at 750 C 2 3 Mg(g) 97.86% MgO 39.10% Ca 1.77% CaO 56.20% (g) Mn(g) 0.20% Al2O3 0.48% Silicothermic reduction SiO 0.13% (g) Metal 227 kg FeSi 227 kg o Mg 0.02% Fe 25% at 1750 C and 85 kPa 2(g) Mg 96.16% Fe (g) 0.02% Si 75% CaMg2 3.14% Slag 1190 kg Al (g) 23.9 ppm Ca2Si 0.31% Aluminium 18.5 kg MgO 6.09% Si(g) 1.60 ppm CaO 0.16% Fe 0.15% CaO 59.76% Al O 1.71 ppm 2 (g) Fe_bcc 0.02% Si 0.06% Al2O3 9.42% Al 99.75% SiO2 24.73% CaAl2 43.1 ppm Mn 0.002% FeSi 170 kg CaO 41 kg Figure 6: Equilibrium calculation for the Mintek process at 1750oC and 85 kPa.

The results of the multistage equilibrium modelling from the present study for the case of Pidgeon, Magnetherm and Mintek process are shown in Figures 7, 8, and 9, respectively. The mass fraction of solids precipitated during each equilibrium stage was plotted against the decreasing temperature. In the case of Pidgeon process, between 1100oC and 900oC, the major solid phases predicted to pre- cipitate include CaO, Fe and FeSi; while various intermetallics were predicted to precipitate below 900oC, as shown in Figure 7. At temperature range between 900oC and 600oC, it was predicted that Mg and Ca are in equilibrium in the vapour phase, thus there was no solid precipitated at such range of temperature. The Ca vapour began to precipitate and interact with magnesium forming CaMg2 at o 600 C, and magnesium was predicted to condense with the impurity of CaMg2.

As can be seen in Figure 8, the major solid phase predicted to precipitate between 1500oC and 1300oC in the case of Magnetherm process were calcium orthosilicate and FeSi. At lower tempera- ture, between 1250 and 1050oC, the metal impurities vapour was predicted to condense and formed oxide and intermetallic phases. Ca was still the main impurity in the vapour until the Mg metal con- densed. In the case of Mintek process, calcium orthosilicate, FeSi and CaO were predicted to pre- cipitate at temperature range of 1700oC-1600oC. From the model, it was found that intermetallics o o are precipitating below 1600 C, while Mg2SiO4 is precipitating below 1000 C; see Figure 9.

1410 Proceedings of EMC 2009 Distribution of impurities in magnesium production via silicothermic reduction

Figure 7: The mass fraction of solids precipitated from mixture of vapours produced from the Pidgeon process reaction (at 1160oC) at various temperatures step cooling of 50oC calculated from thermodynamic model.

Figure 8: The mass fraction of solids precipitated from mixture of vapours produced from the Magnetherm process reaction (at 1550oC) at various temperatures step cooling of 50oC calculated from current thermodynamic model.

Elemental chemical analyses of the Mg produced from the silicothermic processes showed that the main impurity is Ca [17]. This is understandable as high amount of Ca is present in the reactant materials, i.e. dolomite. Ca vapour is present because of the reaction between CaO in dolomite and silicon, which can be written as:

4CaO + Si = Ca + Ca2SiO4 (6)

Proceedings of EMC 2009 1411 Wulandari, Rhamdhani, Brooks, and Monaghan

The thermodynamic calculations from the present study suggested that calcium is the main impurity (assuming all the precipitated solids end up in magnesium product), which is in agreement with the chemical analysis results. The calcium was mainly present in magnesium metal collected in the form of calcium inter-metallic and calcium oxide [21].

Figure 9: The mass fraction of solids precipitated from mixture of vapours produced from the Mintek process reaction (at 1750oC) at various temperatures step cooling of 50oC calculated from current thermodynamic model. In the present study, the equilibrium has been evaluated by: (1) calculating equilibrium after direct cooling of the vapour to the Mg condensation temperatures, i.e. Model I; and (2) conducting multi- stage equilibrium calculations, i.e. Model II. In the Model I, all the impurities in the produced va- pour end up in magnesium. This is not the case in the Model II as the solids condensed at higher temperatures do not advance with the rest of the vapour to the Mg condensation temperatures. Therefore, the total impurities (mainly calcium) in the magnesium calculated using Model I is higher compared to that calculated using Model II. Calcium impurity in the Mintek process pre- dicted by Model II is 0.58%, much less than predicted by Model I, which is 1.77%. Table 4 shows the Mg and the impurities (Ca, Si and Al) concentrations calculated from the present study, as well as the reported actual Ca impurity concentrations [17]. It can be seen from Table 4 that the Ca concentration calculated from Model I is about three to ten times higher compared to that from Model II. It can be seen from the Table 4 that, in the case of Pidgeon and Mintek process, Model I and Model II overestimate the Ca impurity concentration compared to the reported actual Ca concentration [22, 36]. The Ca concentrations from Model I and II are 0.31 and 0.11 wt% for the Pidgeon process model, still much higher compared to the actual data which is 0.02 – 0.06%. How- ever, the result from Model I, in the case of Magnetherm process, lies in the reported actual Ca con- centration range [21].

1412 Proceedings of EMC 2009 Distribution of impurities in magnesium production via silicothermic reduction

Table 4: Magnesium and calcium contents calculated from the present study (Model I and Model II), thermochemical study by Ritter and Sadoway [10], and from chemical analyses [14, 21, 22] Element Pidgeon Magnetherm Mintek

Magnesium (wt %) Model I 99.68 99.94 97.86 Model II 99.89 99.92 99.42 Calcium (wt %) Model I 0.31 0.82 1.77 Model II 0.11 0.08 0.58 Model by [10] - 0.44 – 1.11 - Actual 0.02 – 0.06 0.77-1.05 0.03 – 0.385 Silicon (wt %) Model I 0.002 0.11 0.08 Model II N.A N.A. N.A Model by [10] - 0.44 - 1.19 - Actual 0.03 – 0.07 0.11 – 0.16 0.1 – 1.1 Aluminium (wt %) Model I 0.0005 0.00256 0.00247 Model II N.A N.A N.A. Model by [10] - 0.0012 – 0.0034 - Actual 0.007 – 0.02 0.037 – 0.088 0.01 – 0.066 The result for the Magnetherm process is also compared to one specific study of analysis of impuri- ties in the Magnetherm process by Ritter and Sadoway [10] using thermochemical databases. For calcium impurities, the Model I and II results agreed with their results. The results presented provide information on the behaviour of species during condensation of va- pour produced from silicothermic processes determined from chemical equilibrium reactions. In the actual process, the condensation of magnesium vapour is a very complex phenomena involving fluid and kinetics effects rather than idealized equilibrium process. The vapour may condense ho- mogeneously in the vapour space or via heterogeneous nucleation on particles in the gas phase or along the walls of the condenser. The kinetics of cooling, connected to the heat transfer characteris- tics of the condenser and the fluid mechanics of the system, will affect the rate of nucleation of condensation. Other chemical interactions which are not associated with the main reactions, such as interaction of magnesium with air in the Magnetherm process, and the interaction of carbon lining and electrode with magnesium oxide are also not reflected in the models developed in this study. The equilibrium models developed in this study do not give information about kinetic phenomena but can identify the limits of the process and provide a starting point for more detailed studies of this system.

5 Conclusion

In the present article, a review of three silicothermic Mg productions and thermodynamic analyses of the processes with the emphasis on the formation of impurities and its distribution during con-

Proceedings of EMC 2009 1413 Wulandari, Rhamdhani, Brooks, and Monaghan densation of Mg vapour have been presented. One-stage direct cooling and multi-stage equilibrium modelling have also been presented. The results from the multi-stage equilibrium model show that at equilibrium it should be possible to condense solid impurities before the majority of the magne- sium vapour is condensed below 750oC, depending on the operating pressure of the system and the particular chemistry of the process. The model predicts that purities above 99.8wt% and 99.4wt% can be achieved for the Pidgeon and Mintek process, respectively. Without multi-stage condensa- tion, the impurities in the metal are predicted to be significantly higher for the Mintek process (2.14 wt% compared to 0.6 wt%). These results are broadly consistent with existing industrial data.

6 References

[1] BROOKS, G., COOKSEY, M., WELLWOOD, G., and GOODES, C.: Challenges in light metals pro- duction. Mineral Processing and Extractive Metallurgy (Trans IMM C), 116, (2007), pp. 25-33. [2] ZANG, J.C. and DING, W.: The pidgeon process in china and its future. Warrendale, PA; Magne- sium technology, (2001), pp. 7-10. [3] PIDGEON, L.M. and ALEXANDER, W.A.: Thermal production of magnesium -- pilot-plant studies on the retort ferrosilicon process. Transaction of the AIME. 159 (1944), pp. 315-351. [4] MAYER, A.: Plant for production of magnesium by the ferrosilicon process. Trans. AIME. 159 (1944), pp. 363-376. [5] FAURE, C. and MARCHAL, J.: Magnesium by the magnetherm process. Journal of Metals. (1964), pp. 721-723. [6] CHRISTINI, R.A.: Equlibria among metals, slag, and gas phases in the magnetherm process. Light Metals. (1980), pp. 982-995. [7] CAMERON, A.M., LEWIS, L.A., and DRUMM, C.F.: The thermodynamic and economic modelling of a novel Mg production process. Manchester, England; Proceeding of the 3rd International Magnesium Conference, (1996), pp. 7-18. [8] SCHOUKENS, A.F.S.: A plasma-arc process for the production of magnesium. London, UK; Ex- tractive Metallurgy'89, (1989) pp. 209-223. [9] SCHOUKENS, A.F.S., ABDEL-LATIF, M.A., and FREEMAN, M.J.: Technological breakthrough of the mintek thermal Mg process. The Journal of South African Institute of Mining and Metal- lurgy. 106 (2006), pp. 25-29. [10] RITTER, C.S. and SADOWAY, D.R.: A thermochemical study of the behavior of impurities in the magnetherm process. Phoenix, Arizona; USA; Light Metals, (1988), pp. 799-805. [11] CAMERON, A.A., LOTENS, J.P., and AURICH, V.G.: Carbothermic production of magnesium. London, England; Pyrometallurgy'87, (1987), pp. 195-222. [12] BROOKS, G., NAGLE, M., TASSIOS, S., and TRANG, S.: The physical chemistry of the carbother- mic route to magnesium. Warrendale PA; Magnesium Technology, (2006), pp. 25-33. [13] KIPOUROS, G.J. and SADOWAY, D.R., in Advances in molten salt chemistry. (1987). p. 127-209. [14] TOGURI, J.M. and PIDGEON, L.M.: High temperature studies of metallurgical processes. Part II. The thermal reduction of calcined dolomite with silicon. Canadian Journal of Chemistry. 40 (1962), pp. 1769-1776. [15] PIDGEON, L.M. and KING, J.A.: The vapor pressure of magnesium in the thermal reduction of mgo by ferrosilicon. Discussion of Faraday Society. 4 (1948), pp. 197-206. [16] GALOVSKY, U. and KUHLEIN, M.: A new conti-process for the fluxless recycling of high purity magnesium. Magnesium Technology, (2001), pp. 49-53.

1414 Proceedings of EMC 2009 Distribution of impurities in magnesium production via silicothermic reduction

[17] FROATS, A.: Pidgeon silicothermic process in the 70's. Light Metals, (1970), pp. 969-979. [18] BROOKS, G., TRANG, S., WITT, P., KHAN, M.N.H., and NAGLE, M.: Carbothermic route to mag- nesium. Journal of Minerals, Metals and Materials Society. (2006), pp. 51-55. [19] CHRISTINI, R.A. and BALLAIN, M.D.: Magnetherm atmospheric pressure operation: Aluminum reactivity in silicate slag. Light Metals, Proceedings of 120th TMS Annual Meeting, TMS Light Metals Committee, New Orleans, Louisiana. (1991), pp. 1189-1196. [20] TROCME, F.: The development of the magnetherm process. Light Metals, 1971, pp. 669-678. [21] BOWMAN, K.A.: Magnesium by the magnetherm process -- process contamination and fused salt refining. Light Metals. 2 (1986), pp. 1033-1038. [22] ABDEL-LATIF, M.A.: Refining testwork on crude magnesium produced in the mintek thermal magnesium process. Johannesburg; Southern African Pyrometallurgy (2006), pp. 343-355. [23] HABASHI, F., Magnesium, in Handbook of extractive metallurgy. 1997, Wiley-VCH: Winheim. p. 981-1038. [24] MARSHALL, M., LIU, Z.-K., and CHRISTINI, R.: A computational thermodynamic analysis of at- mospheric magnesium pressure. Magnesium Technology, (2001), pp. 17-20. [25] BARCZA, N.A. and SCHOUKENS, A.F.S., US patent 4,699,653 thermal production of magnesium. 1987, Council for Mineral Technology, Johannesburg, South Africa: South Africa. p. 1-5. [26] ABDEL-LATIF, M.A.: Mintek thermal magnesium process (mtmp): Theoretical and operational aspects. Johannesburg; Southern African Pyrometallurgy 2006, pp. 329-331. [27] WYNNYCKYJ, J.R. and PIDGEON, L.M.: Equilibria in the silicothermic reduction of calcined dolomite. Metallurgical Transactions. 2 (1971), pp. 979-986. [28] BALE, C.W., CHARTRAND, P., DEGTEROV, S.A., ERIKSSON, G., HACK, K., MAHFOUD, R.B., MELANCON, J., PELTON, A.D., and PETERSEN, S.: FactSage thermochemical software and data- bases. Calphad. 26 (2002), pp. 189-228. [29] Janaf thermochemical tables. 2nd ed, ed. D.R. Stull and H. Prophert. 1971, U.S.: NSRDS. [30] BARIN, I.: Thermochemical data of pure substances. Vol. I 1993, Weinheim Germany: VCH Verlagsgesellschaft mbH. [31] BERMAN, R.G. and BROWN, T.H.: Heat capacity of minerals in the system Na2O-K2O-CaO- MgO-FeO-Fe2O3-Al2O3-SiO2-TiO2-H2O-CO2: Representation, estimation, and high tempera- ture extrapolation. Contribution to Mineralogy and Petrology. 89 (1985), pp. 168-183. [32] ERIKSSON, G., WU, P., and PELTON, A.D.: Critical evaluation and optimization of the thermody- namic properties and phase diagram of the MgO-Al2O3, MnO-Al2O3, FeO-Al2O3, Na2O-Al2O3, and K2O-Al2O3 systems. Calphad. 17 (1993), pp. 189-205. [33] TOGURI, J.M. and PIDGEON, L.M.: High temperature studies of metallurgical processes. Part I. Thermal reduction of MgO with Si. Canadian Journal of Chemistry. 39 (1961), pp. 540-547. [34] BETCHERMAN, I.I. and PIDGEON, L.M.: The physical nature of solid condensates produced in the distillation of volatile metals. Transaction of Canadian Institute of Mining and Metallurgy. 54 (1951), pp. 460-465. [35] MISRA, R.N., SAMPATH, V.S., and BHATNAGAR, P.P.: Studies on silicothermic reduction of dolomite for the production of Mg. Transaction of the Indian Institute of Metals. 17 (1964), pp. 145-154. [36] YUCEL, O., YIGIT, S., and DERIN, B.: Production of Mg from turkish calcined dolomite using vacuum silicothermic reduction method. Materials Science Forum. 488-489 (2005), pp. 39-42.

Proceedings of EMC 2009 1415